Impact of livestock interventions on maternal and child nutrition outcomes in Africa: A systematic review and meta-analysis protocol

The challenge of undernutrition (stunting and wasting) still remains a major health concern in children below 5 years of age in Africa, with the continent accounting for more than one third of all stunted children and more than one quarter of all wasted children globally. Despite the growing evidence on the role of agriculture interventions in improving nutrition, empirical evidence on the impact of livestock intervention on nutrition in Africa is scant. This review is aimed at determining whether livestock interventions are effective in reducing undernutrition in children below five years of age and in pregnant and lactating women in Africa. The review will be conducted according to PRISMA guidelines. Major electronic databases will be searched and complemented with grey and non-indexed literature from google and google scholar, and expert consultation for additional articles and reports. PICO criteria will be used while employing search strategies including MeSH, Boolean search operators and truncation/wildcard symbol to narrow or broaden the search. Articles on effect of livestock interventions on maternal and child nutrition conducted in Africa that meet the set inclusion criteria will be included in the review after critical appraisal by two independent reviewers. A standardized form will be used to extract data from included studies. The extracted data will be summarized and synthesized both qualitatively and quantitatively and key outcomes presented. Evidence generated from the systematic review and meta-analysis will be important for guiding nutrition sensitive livestock interventions and policies on nutrition programming, specifically on how to leverage on livestock interventions to reduce the burden of undernutrition.


Introduction
Undernutrition (stunting, wasting and underweight) remains a serious public health concern in Africa 1 . Globally, the continent accounted for more than one third of stunted children (58.7 million, 39%) and more than one quarter of wasted children (13.8 million, 27%) in 2018 2, 3 . To combat and address the challenge of undernutrition and contribute to the progress of attaining Sustainable Development Goal 2 of ending hunger and all forms of malnutrition, holistic, multifaceted strategies employing both nutrition specific and nutrition sensitive interventions are required 4,5 . Additionally, more efforts are needed in building more resilient, equitable and sustainable food systems for improved nutrition 6 Agriculture including livestock plays a key role as a source of food and nutrition security and livelihoods for a majority of rural households in sub-Saharan Africa 7 . In the last two decades, several reviews have been conducted to assess the contribution/impact of general agriculture interventions (home gardening for fruits and vegetables, aquaculture, livestock production, cash crops and biofortified crops) on nutrition 8-18 . These reviews have documented the growing evidence on the role of agriculture interventions in improving nutrition and identified some of the pathways through which agriculture interventions can contribute to nutrition. Animal-source foods (ASF) are a rich source of bioavailable nutrients that play an important role in reducing risk of protein malnutrition 19,20 . In the context of arid and semi-arid areas with limited potential for crop agriculture, the role of livestock and ASF in supporting the livelihood and nutrition of pastoralist communities is especially critical.
The impact pathways through which livestock interventions may influence human nutrition include: (1) Increased production and consumption of animal source foods and hence dietary diversity at household and individual level (2) Increased household level income through sale of livestock products which in turn translates into increased access to dietary diversity 18,21 . Livestock interventions such as dairy programs, small livestock rearing, backyard poultry production, breed improvement, fisheries, livestock transfer programs, livestock feeds improvement and livestock value chains programs have a potential to positively influence improved dietary diversity at household level and possibly impacting the individual nutritional outcomes. However, empirical data on the net contribution of livestock intervention on nutrition in Africa is scant.
This review is aimed at collating, synthesizing and documenting all available evidence on the linkages between livestock interventions and nutrition outcomes in Africa. Evidence generated from the systematic review and meta-analysis will be important for guiding nutrition sensitive livestock interventions and policies on nutrition programming, specifically on how to leverage on livestock interventions to reduce the burden of undernutrition.

Review question
Are livestock interventions effective in reducing undernutrition in children below five years of age and in pregnant and lactating women in Africa?
Objectives of this review are to 1. Assess the available evidence on impact of livestock interventions on maternal and child nutrition outcomes in Africa and identify data gaps 2. Determine the characteristics (design and implementation strategies) of livestock interventions that improve nutrition outcomes 3. Estimate the pooled effect of livestock interventions to improve nutrition outcomes 4. Evaluate the type of livestock interventions more effective in improving nutrition outcomes

Methods
The systematic review will be conducted following the guidelines suggested in the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analysis) statement 22,23 . The review protocol has been registered in the international prospective register of systematic reviews (PROSPERO), protocol registration number CRD42020203843 on 14 th September 2020.

Definitions
Livestock interventions -all livestock related interventions or programmes with an objective of increasing production diversity, access and consumption of animal source foods (ASFs) and income generation to the households. Such interventions include provision of livestock feed, provision of animal health care, provision of water, provision of shelter, and training/extension services

Amendments from Version 1
In the abstract and introduction, we have included underweight as part of the nutrition outcomes to be considered and defined PRISMA and PICO in full. Our second objective has been clarified that the characteristics of interventions to be considered are design and implementation strategies. In the literature search and study selection, we have excluded grey literature and included MeSH terms as part of the search strategy. Under the inclusion and exclusion criteria, we have clarified that studies published any date up to the time of conducting the search will be eligible. Under the data management, we have indicated that duplicates will be removed using the Rayyan QCRI web and mobile application. Additionally for the screening process we have clarified that two independent reviewers will examine the studies and any disagreements among the reviewers on eligibility of any study will be decided through discussions with a third reviewer. For the study designs, we have clarified that experimental studies will include controlled trials while observational studies will include cohort studies, case control studies and cross-sectional studies. For consistency we have clarified our study population will be children under five years and pregnant and lactating women. For data analysis we have indicated that sub-group analysis will be performed to check for check for the sources of heterogeneity and if heterogeneity is observed even after sub-group analysis, then a meta-regression will be conducted.

REVISED
Livestock -all domesticated animals such as cattle, camels, goats, sheep, pigs, other small ruminants, poultry/chicken, fish and bees

Literature search and study selection
Major electronic databases including PubMed, Scopus and Web of Science will be searched by two independent reviewers to identify relevant peer-reviewed publications and reports. All the reference lists of all papers identified through the database searches and relevant papers and reports considered will be reviewed and "forward citation" tool in Google Scholar will be applied to find papers that cited these studies to complement the search. Reference lists of previous systematic reviews conducted on similar study themes will also be reviewed. Experts in this field and study investigators will also be consulted for any additional papers or reports which may not have been captured through the online search.
Search strategy will be based on key words formulated according to the population/patient/problem, intervention/indicator/ exposure, comparison/control, outcome (PICO) format. These key words will be generated through a preliminary general search in major electronic databases to identify most used key words in the publications. Medical Subject Headings (MeSH) terms will be used to identify potential key wards and choose appropriate terms. Boolean operators' terms "AND", "OR" and "NOT" will be used to connect the search terms to either narrow or broaden the search. Truncation/wildcard symbol (*) will also be used for words where variations may be possible (Table 1).

Inclusion and exclusion criteria
Studies will be screened against a set inclusion and exclusion criteria to determine and assess their relevance for inclusion in the systematic review (Table 2). Studies published in any date up to the time of conducting the search and which meet the inclusion criteria will be eligible for inclusion.

Data management
The search results will be uploaded to Rayyan QCRI a web and mobile app for systematic reviews that facilitates collaboration among reviewers during the study selection process. All duplicates will be removed prior to screening using the Rayyan QCRI platform. Screening questions will be developed and tested based on the inclusion and exclusion criteria prior to the start of the screening process.
A two-stage screening process will be employed in all the retrieved articles from the database searches; i. Titles/abstracts will be screened by two independent reviewers for relevance to the review question ii. Full texts of possible relevant articles will be reviewed by two independent reviewers to ascertain if the methods used in the studies selected at stage one adheres to the set methodological standards for the review and exclude those that do not meet the criteria. Any disagreement between the two reviewers over the eligibility of any study will be decided through discussion with a third reviewer and consensus reached.

Data abstraction and synthesis
For those articles found relevant after full text review, data will be extracted using a pre-prepared excel spreadsheet template (extended data 24 ). Variables to be extracted are described in Table 3 The data extracted will be summarized and synthesized both qualitatively and quantitatively and key outcomes presented. For   the qualitative data, a summary of key outcomes will be provided using Excel. For the quantitative synthesis statistical software RevMan 5.1 will be used. The primary outcome measure will be the nutrition status of children below five years and pregnant and lactating women and will be measured through micronutrient status or anthropometry (stunting, wasting and underweight). The intermediate outcome measure will be dietary diversity, incomes and morbidity.
Homogeneity in reporting metrics of the included studies will be assessed. If a sufficient number of studies reporting on effect of livestock interventions on nutrition are identified and there is consistency in reporting metrics, a meta-analysis will be conducted on this subset of studies. Pooled/summary effect estimate of livestock interventions will be calculated using relative risk ratios (RR) and their corresponding 95% confidence intervals (CI). The percentage of variations across the studies and their impact on the meta-analysis will be quantified by calculating and reporting the statistical measure of heterogeneity (I 2 statistic) and obtaining a summary estimate of the effect of livestock interventions on nutrition. For the studies included in the quantitative synthesis, a fixed and random effect models will be used to calculate the RR and 95% CI based on the level of heterogeneity (I 2 statistic) 25,26 . If high levels of heterogeneity are detected ((I 2 >=50% or P <0.1) we will perform a sub-group analysis to determine and explain the source of heterogeneity. If heterogeneity is observed in the sub-group analysis, a meta-regression will be conducted.

Validity/risk of bias assessment
Individual studies will be assessed for both internal and external validity. The Grades of Recommendations, Assessment, Development and Evaluation (GRADE) guidelines 27 will be used to assess study validity/risk of bias. Studies will be scored as either low, medium and high quality based on five criteria; counterfactual analysis, sample size and power calculations, nutrition outcome assessment, intermediate outcome assessment and confounding bias assessment. Overall assessment of risk of bias for each study will be determined through weighted judgement of the established criteria.

Presentation of results
Data summarized and synthesized using qualitative methods will be presented inform of summary tables of key outcomes together with a narrative description of the studies using excel.
For the quantitative analysis (meta-analysis) statistical software RevMan 5.1 will be used for analysis and the output will be presented graphically using a forest plot indicating point estimate and 95% confidence interval of observed effect for each individual study together with summary estimate and its confidence interval.

Dissemination of information
The review findings will be disseminated through open access publication of the results, as well as dissemination in seminars and workshops

Study status
We are currently piloting the study selection process

Willy Kiboi
School of Nursing and Public Health, Chuka University, Chuka, Kenya The paper is generally well written. The methodology used is appropriate but requires some improvements. The following comments may help improve the paper; In the abstract and the introduction sections, the authors may wish to explain why they are excluding underweight children, when in the methodology they state that they will be considered (Table 2).

○
Let the authors write all abbreviations in full in their first use.

○
The authors talk about maternal and child nutrition outcomes in the topic, and yet the aspect of maternal nutrition (pregnant and lactating mothers) is relatively silent in the whole paper.

○
In the second paragraph of the introduction section, the authors mention that several reviews have been conducted to assess the contribution/impact of general agriculture interventions including livestock production on nutrition. Could the authors explain why the review will be necessary if such data exists?

○
The authors may consider explaining further how the study selection will be conducted e.g. how will duplicate and irrelevant articles be excluded.

○
In the study designs to be included, the authors may wish to redefine the designs. For example, quasi-experimental designs are part of experimental studies. I also find the term "observational studies" to be a broad term when explaining a research design. I suggest that the authors be more specific on the research designs that they wish to include in their review.

○
Below Table 2 is the word MUAC (mid-upper arm circumference) which seems to be ○ misplaced. If it is an explanation of the abbreviation, then the authors may wish to have a section with a list of all used abbreviations and their explanations.
In the data management section, the authors state that only articles where there is concurrence between the two reviewers will be included in the review. Inconsistent with that statement, the subsequent statement says that the disagreement between the two reviewers will be discussed with a third reviewer. The researchers may wish to relook at those statements.
○ Furthermore, the authors mention that the primary outcome measure will be the nutrition status of children below five years and women of childbearing age. I find this to have departed from the earlier stated study population (pregnant and lactating women). Women of reproductive age may include all women aged 15-49 as defined by WHO.

○
In the study's heterogeneity assessment, the authors have some contradictory statements. For example, the authors state that "if high levels of heterogeneity are detected, we will perform a subgroup analysis to determine and explain the sources of heterogeneity". In the subsequent statement the authors state that "If heterogeneity is substantial, a metaanalysis will not be performed".
Thank you for your insightful comments. We have carefully addressed each of the comments as detailed below.
In the abstract and the introduction sections, the authors may wish to explain why they are excluding underweight children, when in the methodology they state that they will be considered ( Table 2). This has been reviewed and underweight children has been included The authors talk about maternal and child nutrition outcomes in the topic, and yet the aspect of maternal nutrition (pregnant and lactating mothers) is relatively silent in the whole paper. This has been clarified. Studies evaluating maternal nutrition outcomes for pregnant and lactating mothers will be included in the review.

○
In the second paragraph of the introduction section, the authors mention that several reviews have been conducted to assess the contribution/impact of general agriculture interventions including livestock production on nutrition. Could the authors explain why the review will be necessary if such data exists?
Most of the reviews have focused on crop agriculture. The few livestock reviews have not been done specifically for Africa and have not focused on their impact on nutrition for children under five years and pregnant and lactating women. Our review is aimed at providing evidence for the impact of livestock interventions on nutrition outcomes for children under five years and pregnant and lactating women who are among the most nutritionally vulnerable groups specifically in Africa setting.

In the objectives of the review, it's not clear what the authors mean by characteristics of livestock interventions (objective 2).
This has been clarified, it will be in terms of design and implementation strategies.

○
The authors may consider explaining further how the study selection will be conducted e.g. how will duplicate and irrelevant articles be excluded. This has been clarified. The duplicates and irrelevant articles will be excluded using the Rayyan QCRI application which is a web and mobile phone app for systematic reviews In the data management section, the authors state that only articles where there is concurrence between the two reviewers will be included in the review. Inconsistent with that statement, the subsequent statement says that the disagreement between the two reviewers will be discussed with a third reviewer. The researchers may wish to relook at those statements. This has been clarified. Studies will be examined independently by two reviewers. Studies that meet inclusion criteria will pass for full text review for eligibility by two reviewers. Any disagreement between them over the eligibility of any study will be decided through discussion with a third reviewer. This has been clarified and redundancies in the statement removed. Sub-group analysis will be performed for heterogeneity assessment.
help identify all the potential keywords and will reduce the chance of missing.
Although it was mentioned in table 2 that any publication date will be considered, it is better to cite that in the method write up.

2.
Table 2 is contradictory to the method write-up in terms of literature exclusion criteria. In the table exclusion criteria it was mentioned that unpublished reports will be excluded while in the method section it was mentioned that grey literature will be included. I suggest to clarify the inconsistencies.
For selecting keywords for literature search, I would suggest including MeSH terms. This will help identify all the potential keywords and will reduce the chance of missing.
We agree with the reviewer and Medical Subject Headings (MeSH) has been included as part of the search strategy.

○
Although it was mentioned in table 2 that any publication date will be considered, it is better to cite that in the method write up This has been included in the methods as part of the inclusion criteria.
○ Table 2 is contradictory to the method write-up in terms of literature exclusion criteria. In the table exclusion criteria, it was mentioned that unpublished reports will be excluded while in the method section it was mentioned that grey literature will be included. I suggest clarifying the inconsistencies.
We agree with the reviewer, and for consistency we have deleted the section in the text indicating inclusion of grey literature.
○ Table 2: mentioned that feasibility studies will be excluded from this review. Feasibility studies encompass a wide range starting from pilot projects to effectiveness studies. What was the rationale behind this exclusion? Majority of pilot and feasibility studies are non-randomized and do not include power calculations to test hypothesis. However, some are conducted within standards hence could provide vital evidence. Consequently, in this review feasibility studies will be evaluated against the set criteria and those that meet the set criteria will be included ○ There is mention about using multiple search engines. In that case, the first step will be the identification and removal of duplicate literature. I suggest including duplicate identification and removal methods. Duplicates will be removed using the Rayyan QCRI web and mobile app for systematic reviews.